Turns the admin panel into a real monitoring tool so production bugs are
visible instead of silent.
- Error & crash tracking: error_events table (migration 0010) + logError()
helper + /api/errors ingest; global-error.tsx and (app)/error.tsx report
crashes automatically; /admin/errors viewer (recent + grouped, filters).
- Full audit-log viewer at /admin/audit over the existing audit_log (all
actions, not just auth) with action/resource/family/user/text filters.
- AI observability at /admin/ai over ai_usage: per-intent latency (avg/p95),
tokens, cost, daily trend, slowest calls, medical-redirect count.
- System health at /admin/health: DB latency, migration status, recent error
volume, and integration config presence.
- Sidebar updated with Health / Errors / Audit Log / AI Usage.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
